HIGH SALIENCE / GLOSSARY

REFERENCE · SEPTEMBER 26, 2026

AI Search Glossary

Plain definitions of the terms used in AI search and AI SEO, from AEO and GEO to query fan-out and recommendation share. Each entry is short on purpose and links to a fuller guide where one exists.

BY MIKE HAWLEY, FOUNDER · UPDATED SEPTEMBER 26, 2026

AEO (answer engine optimization)

Optimizing content and brand presence so you are used as the answer in featured snippets, voice responses and AI-generated answers. See AEO vs SEO and our AEO service.

AI systems that carry out several steps for a user, such as searching, comparing options and opening pages, instead of returning a single answer. Still early, and not yet something anyone can measure reliably for brand visibility.

AI citation

A source an AI answer links to or names as its evidence. Citations are the one part of how an answer is built that you can observe from the outside. See AI citations.

AI Mode

Google's conversational search experience, which breaks a question into many searches (query fan-out) and writes one answer with links.

AI Overviews

The AI-written summaries Google shows at the top of some search results, with links to sources. Google says they rely on its core ranking systems.

AI SEO

The umbrella term for optimizing across Google and AI search as one program. See AI SEO.

AI visibility

How often and how favorably a brand appears in AI answers to relevant questions. Best reported as presence, recommendation share and citation share. See AI visibility.

AIO

An ambiguous abbreviation. Some use it for AI optimization in general, others for Google's AI Overviews. Ask which is meant.

Alternatives query

A buyer question such as "X alternatives" or "cheaper alternative to X". The answers show which brands an AI system treats as substitutes for X. Our brand pages report what ChatGPT suggests instead of each brand.

Anchor mention

In our research codebook, a brand named in the question itself, such as HubSpot in "HubSpot alternatives". It is counted separately so the question does not inflate the brand's numbers.

Answer engine

Any system that responds with an answer rather than a list of links: AI chat products, AI search features and voice assistants.

Bingbot

Microsoft's search crawler. Bing's index is what Microsoft Copilot's web answers are grounded in, so a site Bing cannot crawl is hard for Copilot to cite.

Brand mention

Any appearance of a brand's name on a page or in an answer, with or without a link.

Brand salience

A marketing-science term for how easily a brand comes to mind in buying situations, associated with the Ehrenberg-Bass Institute. It is the idea underneath Category Salience.

Category Salience

High Salience's name for how hard a brand is to avoid across the questions buyers ask in its category, in Google results and AI answers together. See the methodology.

The web search mode in ChatGPT, launched in October 2024, which answers with live sources and citations. OpenAI says sites that want to appear in it should allow OAI-SearchBot.

ChatGPT-User

OpenAI's agent for fetching a page when a ChatGPT user asks for it. It is user-initiated rather than automatic crawling.

Chunking

Splitting text into small passages so a retrieval system can match them to a question. Retrieval systems do this themselves; Google says sites do not need to restructure content into chunks for its AI features.

Citation share

How often your own pages appear among the sources AI answers cite, for a defined set of questions.

Claude-SearchBot

Anthropic's crawler for improving Claude's search results. Allow it if you want to appear in Claude search answers.

ClaudeBot

Anthropic's crawler for collecting model training data. Blocking it is a training decision, separate from search visibility. See AI crawlers.

Co-occurrence

Two brands appearing in the same answer or on the same page. In AI answers, the brands that co-occur most often with yours are the competitive set the model has built for you, which is not always the one you would choose.

Comparison query

A head-to-head buyer question such as "Pipedrive vs HubSpot". In our teardowns ChatGPT usually recommends both options for different buyers rather than naming a winner.

Context window

The amount of text a language model can take into account at once, including the question, any retrieved pages and its own answer so far.

Searching by asking questions in natural language, with follow-ups that keep the earlier context, as in ChatGPT, Perplexity and Google AI Mode.

Copilot

Microsoft's AI assistant. Its answers about the web are grounded in Bing search results.

Digital PR

Earning coverage and mentions on news, trade and review sites. For AI search its value is that those sites are among the sources models read when they describe a category. See source authority.

E-E-A-T

Experience, expertise, authoritativeness and trustworthiness: the framework Google's human quality raters use to judge pages. The first E, for experience, was added in December 2022. It is a guideline for raters, not a single ranking score.

Embedding

A list of numbers that represents the meaning of a piece of text, so that passages with similar meaning sit close together. Embeddings are how many retrieval systems find passages that match a question.

Entity

A distinct thing a search system recognizes, such as a company, product or person, with attributes and relationships. Consistent entity information helps systems describe you correctly. See semantic SEO.

Entity salience

In natural language processing, a score for how central an entity is to a document. Google's Cloud Natural Language API reports it on a scale from 0 to 1.

The boxed answer Google shows above regular results for some questions. The original target of AEO.

Gemini

Google's family of AI models and the name of its AI assistant app.

Generative AI

AI systems that produce new text, images or other content in response to a prompt, rather than only classifying or retrieving existing content.

GEO (generative engine optimization)

Improving how often and how favorably a brand appears in answers written by generative AI systems. Named in a 2023 research paper. See GEO vs SEO and our GEO service.

Google-Extended

A robots.txt control that governs whether content is used for Google's Gemini models. It does not affect inclusion or ranking in Google Search. See AI crawlers.

Googlebot

Google's search crawler. It covers Google Search, including AI Overviews and AI Mode, so blocking it removes a site from all of them.

GPTBot

OpenAI's crawler for model training. Blocking it does not remove a site from ChatGPT search, which uses OAI-SearchBot.

Grounding

Tying an AI answer to retrieved sources rather than to the model's training data alone, which is what makes citations possible.

Hallucination

An AI answer that states something false with confidence, such as a wrong price or a feature a product does not have.

Knowledge cutoff

The date after which a model's training data ends. Anything newer can only reach an answer through retrieval, which is why AI search products fetch live pages.

Knowledge Graph

Google's database of entities and the facts connecting them, introduced in 2012. It powers knowledge panels and helps Google recognize what a page is about.

Knowledge panel

The information box Google shows for a recognized entity, such as a company or person, drawn from its Knowledge Graph.

Large language model (LLM)

A model trained on large amounts of text to predict and generate language. ChatGPT, Claude, Gemini and the models behind Perplexity are all built on LLMs.

Listed mention

In our research codebook, a brand an AI answer names as an option without selecting it for any case. The gap between listed and recommended is often the most useful number for a brand.

Listicle

A "best X" list article. Listicles dominate Google's results for many commercial questions and are among the pages AI answers cite, which is why being included in the right ones matters.

LLMO (large language model optimization)

Another name for the same work as GEO and AEO, framed around language models, including how they describe a brand unprompted. See LLM SEO.

llms.txt

A proposed markdown file listing a site's key pages for AI tools. Google Search ignores it, and server logs show major AI crawlers rarely request it. See llms.txt.

OAI-SearchBot

OpenAI's search crawler. OpenAI recommends allowing it in robots.txt for sites that want to appear in ChatGPT search.

Passage ranking

Google's ability to rank a page for a specific passage within it, launched in 2021. Useful context for AI search, since answers are assembled from passages, not whole pages.

People Also Ask

The box of related questions Google shows in many results, each expanding to a short answer and a source. A good map of the questions buyers ask next.

Perplexity

An AI search engine that answers questions with numbered citations to live web sources.

Perplexity-User

Perplexity's agent for fetching a page when a user asks. Perplexity treats it as a user request, so it generally does not follow robots.txt.

PerplexityBot

Perplexity's search crawler, which surfaces and links sites in Perplexity answers. Perplexity says it is not used to train foundation models.

Presence

The share of relevant AI answers that name your brand at all, whether or not they recommend it.

Prompt tracking

Running a fixed set of questions through AI systems on a schedule and recording which brands and sources appear. Answers vary between runs, so results should be read as frequencies over many questions, not as fixed positions.

Query fan-out

Splitting one question into many searches, then combining the results into one answer. Google uses it in AI Mode.

RAG (retrieval-augmented generation)

The pattern most AI search products follow: retrieve relevant documents first, then generate an answer from them.

Recommendation share

Of the AI answers that name a brand, the share that select it for a stated case. The metric closest to a buying decision.

In our research codebook, a brand an AI answer selects for at least one case, such as "best for small teams". See the teardowns for how often each brand gets there.

Retrieval

The step where an AI search system fetches pages or passages relevant to a question before writing its answer. What gets retrieved sets the limits of what the answer can say.

robots.txt

A plain text file at the root of a site that tells crawlers which paths they may fetch. It is how sites allow or block specific AI crawlers. See AI crawlers.

sameAs

A schema.org property that links an entity on your site to its profiles elsewhere, such as LinkedIn or Wikipedia, so systems can connect them as the same company or person.

Search Generative Experience (SGE)

Google's 2023 Search Labs experiment with AI-written answers. It became AI Overviews when Google launched them in the United States in May 2024.

Search intent

What a person wants from a query: information, a specific site, a comparison or a purchase. AI answers and Google results both change shape with intent.

Search that matches meaning rather than exact words, so a question can find a passage that answers it without sharing its phrasing. See semantic SEO.

SERP

Search engine results page: everything Google or another engine shows for a query, including organic results, ads, AI Overviews and other features.

Share of voice

A brand's share of all brand mentions across a defined set of questions, compared with its competitors. In AI search it is usually measured as presence and recommendation share.

Source authority

How much weight search and AI systems give a site as evidence on a topic. In practice, whether the pages that shape answers in your category describe you. See our source authority work.

Structured data

Machine-readable markup, usually schema.org, that describes a page. Useful for rich results and clarity; Google says no special markup is needed for its AI features.

Topical authority

The idea that a site covering a subject thoroughly is treated as a more reliable source on it. Google has described a topic authority system for news; in wider SEO use the term is looser and has no published metric.

Finding the passages whose embeddings are closest to the embedding of a question. It is one of the main retrieval methods behind AI search.

Asking a question aloud to an assistant. Voice answers usually give a single result, which made them an early target of answer engine optimization.

A search that ends without a click to any website, because the answer was on the results page or in the AI response. See zero-click search.

Want the longer explanations? Browse all AI search guides.